COVID-19 – One virus, many faces

Once infection has taken place, COVID-19 becomes a battle at the cellular level; and each individual's cellular response has proved highly heterogeneous. This paper builds on our previous research into excess mortality, taking a deeper dive to help better understand the risks and "many faces" of COVID-19.

Using extensive data mining of electronic health records, researchers have established that age, gender and underlying preconditions are all risk factors that affect the level of illness from COVID-19. Socio-economic factors also have statistical significance in the severity of the disease; as does genetic susceptibility, prior infection history and strength of the immune system. Improved understanding of these vulnerabilities will allow public health authorities to better target resources; will provide re/insurers with greater oversight of their own portfolios; and will support forward looking modelling and contingency planning.  

1. COVID causation and individual response

1.1 The cold war

COVID-19 is a David vs. Goliath battle. The SARS-CoV-2 virus, with its simplicity of four protein types, is waging war against our multi-layered physical and immune defences. David has been remarkably successful; at the time of writing, there were in excess of 20 million global confirmed cases of COVID-19 and more than 730'000 deaths.

There is much we can do to broadly prevent infection, from social distancing and protective measures to developing a vaccine, but once an individual is infected, the battlefield is at a cellular level and the response varies widely. At one end of the spectrum, large numbers of SARS-CoV-2 carriers, ranging from 41%-80% for young adults down to 6%-20% for those in at-risk groups, appear fully asymptomatic1, 2. At the other end, the mortality rate of COVID-19 is a multiple of seasonal influenza, with great variations in the extent of illness in between3.

Age, gender and underlying conditions appear the most significant risk factors in developing COVID-19. A systematic review and meta-analysis of observational studies show that 14 from an initial search of 143 studies attempted to quantify associations between various risk factors and risk of death from COVID-19 infection. The most significant associations were with older age, male gender, hypertension, cardiovascular disease, diabetes, chronic obstructive pulmonary disease and cancer4.

Even within risk clusters, the individual response to viral invasion will be unique, based on genetic susceptibilities, prior history of infections, strength of immune system and many competing positive and negative factors. The emergence of a new virus such as SARS-CoV-2 means that no one will have precisely targeted antibodies. While the details in the science are still being discovered, it is possible that many could have cross-reacting antibodies or T-cell immunity that will limit the likelihood of symptoms and serious infection.

Socio-economic factors further influence the risk of infection. These can vary widely, but are currently thought to include occupation, social deprivation and geography. Another study applied a negative binomial model to COVID-19 deaths in 2,692 counties in the US, after categorising counties into low, medium and high prevalence levels of infection5. Risk factors that were associated with increased death counts in metropolitan areas were, vulnerability of older persons and higher levels of air pollution.

A better understanding of the heterogenous effects of the disease and how individuals respond will improve our prevention measures; will strengthen re/insurers modelling of the progress of the pandemic; will provide stronger foundations to target therapeutic responses; and may hasten progress in finding a vaccine.

At all times, we look to understand the nature of risk. In COVID-19, the risk can be broadly split into three separate, yet linked types:

  1. Risk of infection. This is a measure of the percentage or number of a population infected with SARS-CoV-2 and is known as the infection rate.
  2. Lethality, or risk of death - once infected. The likelihood of those infected dying due to COVID-19, is known as the infection fatality rate (IFR). The risk of dying for those who are infected, and who also have clinical symptoms, is known as the case fatality rate.
  3. Population mortality rate. This is the risk of infection (risk 1) multiplied by the risk of death from COVID-19 if infection with (risk 2) SARS-CoV-2 has occurred.

1.2 The importance of data – electronic Health Records (eHR)

Electronic Health Records (eHR) have been used to evaluate how infection incidence would be affected by travel restrictions and social distancing. The largest study linked primary and secondary care eHR for 3,862,012 men and women registered at general practitioner (GP) practices in England to determine the level of excess deaths for different estimates of R0 (the basic reproduction number is the average number of people infected by one infected person at the start of spread of the virus when everyone in the population is still susceptible to the novel pathogen)6.

However, whilst such studies provide insights on the proportion of individuals in high-risk categories and the effectiveness of different interventions, the data collected was insufficient to determine how infectivity might vary by age, sex and duration post infection. 

Providing more valuable insights, the OpenSafely platform ( was launched on 7 May, using NHS England eHR to look for patterns among hospitalised patients who die from COVID-19. The platform was constructed in five weeks, bringing together 24 million patient primary records into a virtual data centre.

OpenSafely represents a multi-disciplinary collaboration between University of Oxford DataLab, the eHR group at London School of Hygiene and Tropical Medicine, the eHR software developers TPP and wider groups such as ICNARC (Intensive Care National Audit & Research Centre), working on behalf of NHS England and NHSX. The collaboration's first study investigated the mortality rates in hospital with COVID-19 relative to the general population7.

This study was the first to quantify the separate importance of deprivation measures, and confirmed the importance of age, sex and prior medical conditions. In this report we will explore these risk factors in separate chapters, whilst a future report will focus on the impact that COVID-19 has on different organ systems throughout the body. The death hazard ratios are illustrated in Figure 1.1.

Figure 1.1 OpenSafely: estimated hazard ratios for in-hospital deaths primary electronic health records (7)

The OpenSafely project is pioneering a deliberately open approach to both collaboration with leading research institutions and providing access to its open-source code. Approved researchers will be able to carry out large scale cohort analyses within the TPP data centre. Areas of focus could include:

  • Identifying treatments that alter the risk and severity of COVID-19.
  • Identifying individuals at highest risk of hospital admission, ventilation or death to inform government advice and guide planning at all levels.
  • Using local clinical data to predict local spread and health services need.
  • Providing early warnings of disrupted clinical services such as cancer referrals and emergency interventions for heart attacks and stroke to better monitor and measure the likely indirect impact of COVID-19 on population health.

Another database being used by researchers, the UK Biobank, contains medical records, lab samples and survey questionnaires on 502,624 volunteers recruited from 2006 to 2010. Multivariate analysis identified the relative importance of modifiable and non-modifiable risk factors for COVID-19 in the Biobank population8. The modifiable risk factors included higher body mass index, smoking and use of anti-hypertensive medications. The non-modifiable risk factors included older age, male gender, and socio-economic deprivation. The complete multivariate set of relative risks and population attributable fractions from the study are set out in Table 1.1, contrasting pneumonia and COVID-19.

Table 1.1 Comparisons of relative risks and population attributable fractions between COVID-19 and pneumonia from the UK Biobank in the general population (8)

2. Importance of age as a risk factor

2.1 Tracking infections for different age groups

Thanks to extensive contact tracing and widespread testing, South Korea has exemplary quality and quantity of COVID-19 data. Yu, Duan9 analysed confirmed cases and deaths from the Korea Center for Disease Control until early May.

A generalised additive model identified that young people aged 20-39 were infected in the initial outbreak, although this was also linked to the demographics of people attending several places, who were responsible for the initial "super-spreading" events. The rate of infection was significantly lower for those of school age; and both lower and later for those aged 60 and above. Further, infections for those over age 60 continued to fall after 15 March, even though there was a second peak for younger adults in early April. South Korea's rigorous approach to quarantining young people (generally thought to not be at risk) to protect the elderly and those at higher risk, have shown that there is an overall benefit to taking these measures at an early stage. They have demonstrated that elderly people can be effectively protected from virus infection through extensive contact tracing, proper personal protective equipment and early quarantining of those showing symptoms.

2.2 European comparisons

Sobotka, Brzozowska10 obtained data on the distribution of COVID-19 cases across 10 European countries. The quality of data is dependent on the extent and timing of national testing, with considerable variations existing within Europe.

Infection rates were lowest among children and teenagers up to age 20. Several reasons have been given for this, but it is likely that the mild symptoms of COVID-19 often experienced by children are responsible for an unusually low reporting rate. While details around the infectivity of asymptomatic patients is still limited, it is likely that young children and teenagers are not responsible for significant infections requiring hospitalisation.

In contrast to higher rates of male mortality, it would initially appear that there were no general differences in the infection rate between men and women. However, closer examination identified that women of working age are more likely to test positive with COVID-19, whereas there are more confirmed cases among men at older ages (see below).

Figure 2.1 Relative ratio of male to female confirmed COVID-19 cases (10)

The study suggested that gender differences below age 60 were most likely a result of a greater number of women employed in the health and care sectors; and a lack of gender-differentiated personal protective equipment. A further explanatory factor is childcare patterns: in Norway and Germany in particular, women reduce working rates in their 30s and increase again in their 40s.

In terms of population rates, older men also appear more likely to become infected and more likely to die from or with COVID-19 once infected. Comparisons of male and female mortality rates across 6 different countries by Altringer, Zahran11 and as seen in Figure 2.1 demonstrated that the share of COVID-19 in total deaths by age groups held largely constant across Europe, as illustrated in Figure 2.2.

Figure 2.2 Age-specific COVID-19 deaths per million for 6 European countries (11)

Vaupal and others suggest inter-country differences can largely be explained by frailty. Perhaps counterintuitively, countries with factors benefiting longevity also have higher frailty rates and hence are more susceptible to mortality shocks such as COVID-19.

Figures for frailty are based on the European Commission Eurostat database of individuals over age 65 requiring help with household and personal care activities.

Figure 2.3 illustrates the positive association between this frail survival rate and a logarithmic transformation of COVID-19 mortality rates.

Figure 2.3 COVID-19 death rates and health frailty (11)

The study concluded the exponential increase in COVID-19 mortality at older ages and country variations were consistent with a high percentages of deaths taking place in frail populations in care homes, specifically where insufficient zoning meant that elderly, susceptible populations were exposed both to visiting staff and those discharged prematurely from “hotzone” hospitals11.

3. Gender, biochemistry, and the risk of COVID-19

3.1 ACE2 and TMPRSS2

The spike protein of the SARS-CoV-2 virus binds to ACE2 enzymes on the surface of cells to effect entry. The ACE2 enzyme is a key component in the renin-angiotensin pathway that regulates blood pressure. The enzyme converts angiotensin II to angiotensin I (amongst other molecules) and counteracts the effects of angiotensin-converting enzyme (ACE) providing an analogous role to ACE inhibitors such as captopril.

ACE2 is present in cell types in many locations including the lungs, heart, blood vessels, kidneys, liver and gastrointestinal tract. ACE2 is more prevalent at older ages and in patients with hypertension, diabetes and ischaemic heart disease. Angiotensin II increases blood pressure and promotes cell injury. Reduced ACE2 activity because of SARS-CoV-2 infection leads to more inflammation and damage due to higher concentrations of angiotensin II. Damage to surface cells in key organs and in blood vessels increases the likelihood of microthrombi, or small clots, and hence further disruption to blood circulation.

For the virus to enter the cell, the spike protein must be cut by TMPRSS2, a transmembrane protease enzyme. The level of TMPRSS2 expression is dependent on androgenic hormones such as testosterone. One study examined genetic variations in TMPRSS2 in Italians and East Asians to find fewer deleterious mutations and hence was able to predict a higher level of TMPRSS2 activity. These two influences could contribute to a more severe disease course in men12.

One treatment for some prostate cancers is the use of androgen deprivation therapy (ADT) or 5-Alpha reductase inhibitors (SARI), as the cancer is sensitive to concentrations of androgens.  Duga, Asselta13 tracked a cohort of 284 male patients with confirmed COVID-19 and found that the proportion of those taking anti-androgenic therapies was only 4% as compared to 15% in the general population of Italian men over age 40. Although a small sample, this would suggest that anti-androgenic therapies could be inhibiting TMPRSS2 activity, hence reducing the likelihood of COVID-19 infection.

One of the variants predicted to lead to higher TMPRSS2 expression is associated with the single nucleotide polymorphism (SNP) rs2070788. This SNP was previously identified as representing a higher risk in both Influenza A(H7N0) and severe Influenza A(H1N1) from the 2009 pandemic. The former influenza virus was also associated with significantly worse clinical outcomes for men as compared to women.

3.2 Differences in immune response

The two principal categories of antibodies produced by B cells are IgM and IgG. IgM is the largest of the antibody molecules with 10 antigen-binding sites. It is an ideal first responder to a viral infection because of its large size and effective activation of the broader immune response. Later in the course of an infection, B cells will switch to producing IgG, a smaller antibody able to access more locations and which effectively clears pathogens from the blood.

An early study in China followed 331 patients who had experienced different severities of COVID-19 and tracked the levels of IgG antibodies14. Female patients with severe COVID-19 had higher concentrations of IgG antibodies than male patients; and production of IgG antibodies was higher for female patients in the early stages of COVID-19.

These findings are consistent with wider studies that have concluded that the female immune response is stronger and more effective than that of men, in part because a number of key determinant genes are located on the X chromosome. Another early insight was provided through whole exome sequencing of particular ACE2 variants, the protein that SARS-CoV-2 uses to enter cells, that are of low frequency and may be more susceptible to virus cell entry15. The gene for the ACE2 protein is located on the X chromosome, and hence women with their paired copies of the X chromosome would be much less likely to have two copies of these rare variants.

3.3 Matched study between men and women

Using the TriNetX eHR research network, which has 49 million patients, of whom around 50% are in the US, 5,980 men and 7,730 women with confirmed diagnoses of COVID-19, were tracked. Most significantly, they found that up to 94% of overall mortality from COVID-9 was found in patients aged over 5016.

Propensity matching was carried out using a greedy nearest-neighbour matching algorithm. This selects a treated subject and then selects as a matched control subject, the untreated subject whose propensity score is closest to that of the treated subject, allowing for a representative 1:1 comparison. This factored in information such as age, nicotine use and other comorbidities/risk factors such as diabetes, hypertension, chronic lung disease, cardiovascular disease and chronic kidney disease. Relative risks of death (RR 1.4), hospitalisation (RR 1.3) and mechanical ventilation (RR 1.7) were all significantly higher after matching with those diagnosed with COVID-19, confirming that the most serious cases of COVID-19 are likely to require ventilation, but that recovery is possible.

4. Impact of deprivation on COVID-19

Data from over 2 million unique users of the Contact Symptom Tracker app was to examine the spread of COVID-19 across the UK17. Their analysis identified that urban-living, higher deprivation area-level scores and average household size were positively associated with higher COVID-19 prevalence and severity.

It has been further highlighted that there appears to be two different phases in the spread of the virus; and that the relationship with affluence switches between the two phases18. In the initial phase, their German study identified that infection rates were positively associated with the wealth of an area as the virus was spread by business travellers, students or tourists. During the second phase, as governments implemented lockdowns, more deprived areas were less able to reduce interactions and maintain social distancing. Those living in deprived areas were more likely to be vocational workers; to live with vocational workers; or be more exposed to viral load by needing to commute.

Data from building footprints, food access, mobility and transit use in New York City was used to develop a model to test whether differences in infection rates could be explained by an inability to socially isolate or distance19. The study determined that the best predictor of high infection rates was average household size; the proportion of essential workers; and commuters reliant on public transport.

5. Impact of occupation on COVID-19

The UK Office for National Statistics (ONS) produced an interactive tool on 11 May that attempted to assess the risk of exposure for different occupations, as illustrated in Figure 5.1.

A study focused on 120,000 UK Biobank participants who were employed/self-employed at baseline and aged 65 or younger in March 202020. This study identified that those who work in healthcare and social/education were 7.6x and 2.2x  more likely than non-essential workers to test positive in a hospital, although in part this reflects the greater availability of testing. Those in skilled trades, such as electricians or plumbers, and personal service occupations, such hairdressers or beauticians, showed the lowest level of confirmed COVID-19 cases.

Figure 5.1 Probability of positive COVID-19 testing for different occupations (21)

6. Need to protect vulnerable populations

6.1 Importance of knowledge and behaviour to limit the spread  

Behaviours – including willingness/ability to social distance, wearing masks and hygiene measures – have also been found to have a statistically significant impact on increase of COVID-19 risk. A nationally representative survey in the USA in March/April was conducted to assess differences in knowledge, beliefs and behaviour regarding COVID-1922. Questions included identifying top COVID-19 symptoms from a longer list and identifying measures that could best prevent the spread of the virus. At that time, the largest gaps in knowledge appeared to be in men and those under age 55. Interestingly, the survey found that men washed their hands 3.8x fewer every day than women; and those aged 18-29 washed their hands 4.4x fewer each day than older individuals.

6.2 Care homes

Hospitals were much more visible in the early stages of COVID-19, overlooking the plight of care homes. In some countries, personal protective equipment was unavailable; occupancy levels were too high for adequate zoning; care staff routinely rotated between homes; and patients were discharged direct from “hotzone” hospitals, all of which led to devastating outbreaks and chronic staff shortages. Over 45% of the excess deaths in England and Wales so far in 2020 have been care home residents.

For further information, please read our recent report on excess mortality.

6.3 Stratify and shield     

Some authorities have sought to balance protection of vulnerable risk groups with opening the wider economy. These policies were described as “stratify and shield”, in the 26 February meeting of the Scientific Advisory Group for Emergencies (SAGE) -- a group that provides scientific advice to the UK government.  Experience has led most governments to approach easing lockdowns with caution, linking economic opening to contact tracing and clear vigilance of the changes in infectivity over time. This factor, known as Rt, has become an indicator to measure the likely short-term trend of the virus. However, continuing restrictions on particular sectors of the economy may mean that further shifts toward stratify and shield may be necessary.

The number of people who need to protect themselves may be smaller than we currently estimate. T-cell immune responses developed to other prior coronavirus infections may render a larger segment of the population either asymptomatic or experiencing only mild symptoms to COVID-19. Only greater testing can give us a better understanding of this potential cross reactivity, which, while uncertain, is still an area for scientific discovery. Several antibody tests have now been developed, allowing for both faster and more accurate COVID-19 seroprevalence testing.

7. Implications for re/insurers

7.1 Differences between insured and general population

Section 4 highlights evidence that morbidity and mortality experience of COVID-19 is likely to vary across wide groups of people. Since the insured population goes through the underwriting process, they generally exhibit fewer of the comorbidities linked to mortality or severe symptoms of COVID-19, we may expect a lower impact from disease on them, compared to the general population.

The pandemic could have two knock-on effects that will influence re/insurance portfolios:

  • Negative impact on chronic diseases: Interruption to screening programmes, disruption to general medical services and reluctance to attend “hotzone” hospitals may lead to lower numbers of cancers being detected and reduced numbers of heart attacks and strokes presenting as emergencies.
  • Negative impact on mental health conditions: The psychological effects of lockdowns, or on those with underlying conditions continuing to self-isolate, could be considerable. A likely deterioration in economic conditions into 2020 and a rise in unemployment may cause further negative impact on well-being.

7.2 Future mortality and morbidity assumptions

The impact of COVID-19 on future mortality and morbidity assumptions is likely to be complex and has been explored in more detail in our mortality paper. The expectation of significant numbers of excess deaths might be that deaths have been accelerated leading to a survivor population. However, the nature of multi-system organ damage means that new cohorts of frail population have been created, possibly exacerbated by aggressive ventilation treatment in hospitals, which we will address in our next report. Moreover, as mentioned above, there has been a severe interruption in the treatment and diagnosis of chronic conditions, which may have longer term morbidity and mortality implications.

Together, these changes will lead to unexpected disruption in historical trends. They will compel re/insurers to a greater adoption of forward-looking models that consider interactions between risk behaviour, healthcare delivery and external influences. The timing and roll-out of both vaccination programmes and treatments to reduce the onset and severity of COVID-19 symptoms will be likely to promote stochastic models over deterministic scenario approaches.

Contributing editors: Stephen Kramer & Adam Strange    


  1. Davies, N.G., et al., Age-dependent effects in the transmission and control of COVID-19 epidemics. Nat Med, 2020.
  2. World Health Organization. Transmission of SARS-CoV-2: implications for infection prevention precautions. 2020.
  3. Faust, J.S. and C. Del Rio, Assessment of deaths from COVID-19 and from seasonal influenza. JAMA Internal Medicine, 2020.
  4. Parohan, M., et al., Risk factors for mortality in patients with Coronavirus disease 2019 (COVID-19) infection: a systematic review and meta-analysis of observational studies. Aging Male, 2020: p. 1-9.
  5. Tian, T., et al., Risk factors associated with mortality of COVID-19 in 2692 counties of the United States. medRxiv, 2020: p. 2020.05.18.20105544.
  6. Banerjee, A., et al., Estimating excess 1-year mortality associated with the COVID-19 pandemic according to underlying conditions and age: a population-based cohort study. Lancet, 2020. 395(10238): p. 1715-1725.
  7. Williamson, E., et al., OpenSAFELY: factors associated with COVID-19-related hospital death in the linked electronic health records of 17 million adult NHS patients. medRxiv, 2020: p. 2020.05.06.20092999.
  8. Ho, F.K., et al., Modifiable and non-modifiable risk factors for COVID-19: results from UK Biobank. medRxiv, 2020: p. 2020.04.28.20083295.
  9. Yu, X., et al., Distinctive trajectories of COVID-19 epidemic by age and gender: a retrospective modeling of the epidemic in South Korea. Int J Infect Dis, 2020: p. 2020.05.27.20114819.
  10. Sobotka, T., et al., Age, gender and COVID-19 infections. medRxiv, 2020: p. 2020.05.24.20111765.
  11. Altringer, L., S. Zahran, and A. Prasad, The Longevity-Frailty Hypothesis: Evidence from COVID-19 Death Rates in Europe. medRxiv, 2020: p. 2020.04.14.20065540.
  12. Asselta, R., et al., ACE2 and TMPRSS2 variants and expression as candidates to sex and country differences in COVID-19 severity in Italy. Aging (Albany NY), 2020. 12(11): p. 10087-10098.
  13. Duga, S., et al., Impact of anti-androgenic therapies on COVID-19: an observational study in male population from a COVID-19 regional centre of Lombardy (Italy). medRxiv, 2020: p. 2020.04.20.20068056.
  14. Zeng, F., et al., A comparison study of SARS-CoV-2 IgG antibody between male and female COVID-19 patients: a possible reason underlying different outcome between gender. medRxiv, 2020: p. 2020.03.26.20040709.
  15. Benetti, E., et al., ACE2 gene variants may underlie interindividual variability and susceptibility to COVID-19 in the Italian population. medRxiv, 2020: p. 2020.04.03.20047977.
  16. Singh, S., et al., Gender-Based Disparities in COVID-19: Clinical Characteristics and Propensity-matched Analysis of Outcomes. medRxiv, 2020: p. 2020.04.24.20079046.
  17. Bowyer, R., et al., Geo-social gradients in predicted COVID-19 prevalence and severity in Great Britain: results from 2,266,235 users of the COVID-19 Symptoms Tracker app. medRxiv, 2020: p. 2020.04.23.20076521.
  18. Pluemper, T. and E. Neumayer, The COVID-19 Pandemic Predominantly Hits Poor Neighborhoods, or does it? Evidence from Germany. medRxiv, 2020: p. 2020.05.18.20105395.
  19. Carrion, D., et al., Assessing capacity to social distance and neighborhood-level health disparities during the COVID-19 pandemic. medRxiv, 2020: p. 2020.06.02.20120790.
  20. Mutambudzi, M., et al., Occupation and risk of COVID-19: prospective cohort study of 120,621 UK Biobank participants. medRxiv, 2020: p. 2020.05.22.20109892.
  21. Office of National Statistics. Which occupations have the highest potential exposure to the coronavirus (COVID-19)? 2020.
  22. Cutler, D., et al., Disparities in COVID-19 Reported Incidence, Knowledge, and Behavior. medRxiv, 2020: p. 2020.05.15.20095927.